enable-checking failure record_insns () and PARALLEL

[Jeffrey A Law]

  In message < 39048534.D14D67D@rcp.co.uk >you write:
  > 
  > Perhaps record_insns () should be changed to call
  > abort () if it gets passed a PARALLEL.
It would be better to abort if it was passed something that did not have
a UID field.

That would be everything except INSN, JUMP_INSN, CALL_INSN, BARRIER,
CODE_LABEL and NOTE.

If we do this, it should probably be controlled by ENABLE_CHECKING.

jeff